From 0632e3a9413fc00b0f334381d71bef6e1d42e000 Mon Sep 17 00:00:00 2001
From: Chen Meng <chen.meng@intel.com>
Date: Mon, 3 Jun 2024 05:42:40 -0400
Subject: [PATCH] [Layers] Increased the threshold for enabling flashAttn

---
 src/models/env_config.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/models/env_config.cpp b/src/models/env_config.cpp
index 7d3f4928..9c94e9df 100644
--- a/src/models/env_config.cpp
+++ b/src/models/env_config.cpp
@@ -37,7 +37,7 @@ int getFlashThresh() {
     // > threshold to enable flash attention, default 1024
     static int envFlashThresh = -1;
     if (envFlashThresh == -1)
-        envFlashThresh = (getenv("FLASH_ATTN_THRESHOLD") ? atoi(getenv("FLASH_ATTN_THRESHOLD")) : 1024);
+        envFlashThresh = (getenv("FLASH_ATTN_THRESHOLD") ? atoi(getenv("FLASH_ATTN_THRESHOLD")) : 8192);
     return envFlashThresh;
 }